Equilibrium molecular dynamics (EMD) simulations have been applied on a graphical processing unit (GPU)-based computing platform for predicting the thermal conductivity of monolayer graphene. The GPU implementation features calculations of force interactions and heat autocorrelation functions (HACF) using the optimized Tersoff potential. The error associated with GPU-based single-point precision calculations is approximately 1% compared to serial CPU-based double-precision calculations. The calculated thermal conductivity decreases with graphene domain size, while the speedup of the GPU-based computations increases with graphene domain size. These results show that the parallel efficiency of GPU-based calculations for a many-body potential with EMD is low due to the requirement of storing individual potential energy and HACF contributions. The predictions from this study are comparable to those from other studies.

This content is only available via PDF.
You do not currently have access to this content.